Accelerated medical image reconstruction with improved aliasing artifact reduction

Patent metadata

Field	Value
Publication number	US-12374004-B2
Application number	US-202217811889-A
Country	US
Kind code	B2
Filing date	Jul 12, 2022
Priority date	Jul 12, 2022
Publication date	Jul 29, 2025
Grant date	Jul 29, 2025

Abstract

For reconstruction in medical imaging, such as reconstruction in MR imaging, scanning is accelerated by under-sampling. In iterative reconstruction, the input to the regularizer is altered provide for correlation of non-local aliasing artifacts. Duplicates of the input image are shifted by different amounts based on the level of acceleration. The resulting shifted images are used to form the input to the regularizer. Providing an input based on shifts allows the regularization to suppress non-local as well as local aliasing artifacts.

First claim

What is claimed is:

1. A method for reconstruction of a medical image in a medical imaging system, the method comprising:
scanning, by the medical imaging system, a patient with an acceleration factor, the scanning resulting in measurements;
reconstructing, iteratively by an image processor, the medical image from the measurements, the iterative reconstruction including at least a regularizer implemented with one or more machine-learned networks and an alteration stage, wherein for each iteration, the alteration stage generates multiple shifted copies of an input image by applying a circular shift to the input image, the input image provided by an initial image from the measurements or a previous iteration of the iterative reconstruction, wherein the multiple shifted copies of the input image are input to the one or more machine-learned networks implemented by the regularizer and wherein the regularizer is configured to output an output image that is input into a subsequent iteration, wherein the iterative reconstruction proceeds for a plurality of iterations; and
displaying the reconstructed medical image.

2. The method of claim 1 wherein scanning comprises scanning with the medical imaging system being a magnetic resonance (MR) scanner and the measurements being k-space measurements, the scanning comprising parallel imaging combined with compressed sensing.

3. The method of claim 1 wherein reconstructing comprises reconstructing as an unrolled iterative reconstruction where each of multiple reconstructions in the unrolled iterative reconstruction includes regularization.

4. The method of claim 3 wherein reconstructing comprises generating the input as an initialization of the unrolled iterative reconstruction, the regularizer being a first of the multiple reconstructions.

5. The method of claim 1 wherein the alteration stage generates the multiple shifted copies of the input image using the circular shift from an output image of a data-consistency check.

6. The method of claim 1 wherein an amount of the multiple shifted copies substantially equals the acceleration factor.

7. The method of claim 6 wherein forming the multiple shifted copies comprises shifting the input image by different shifts, each of the shifts being a function of the acceleration factor where a number of the multiple shifted copies is provided as the input image and shifted images corresponding to the different shifts.

8. The method of claim 6 wherein forming the multiple shifted copies of the input image comprises shifting the input image by different shifts, the different shifts being fractions with the acceleration factor as a denominator.

9. The method of claim 8 wherein shifting comprises shifting with sequential integers as numerators of the fractions.

10. The method of claim 8 wherein shifting comprises shifting with a field of view of the regularizer multiplied by sequential integers as numerators of the fractions.

11. The method of claim 1 wherein reconstructing comprises inputting a concatenation of the multiple shifted copies of the input image as the input to the regularizer.

12. The method of claim 1 wherein reconstructing further comprises applying convolutional neural networks to the multiple shifted copies of the input image, respectively, the input to the regularizer being a normalization of the multiple shifted copies of the input image output by the convolutional neural networks.

13. The method of claim 12 wherein the normalization comprises a sum and convolution.

14. The method of claim 12 wherein applying the convolutional neural networks comprises applying, independently for each of the multiple shifted copies of the input image, the convolutional neural networks with shared weights.

15. The method of claim 1 wherein a maximum number of channels is provided, and wherein reconstructing comprises forming the multiple shifted copies of the input image where the number is less than the maximum number, wherein zero padding is used for excess ones of the channels.

16. A system for reconstruction in medical imaging, the system comprising:
a medical scanner configured to scan a region of a patient, the scan under sampling by a factor and providing scan data;
an image processor configured to reconstruct a representation of the region from the scan data, the image processor configured to iteratively reconstruct by application of a machine-learned model in a regularization stage, the regularization stage having an input provided by an alteration stage that generates a plurality of circular shifted copies of an input image; and
a display configured to display an image of the region from the reconstructed representation.

17. The system of claim 16 wherein the image processor is configured to reconstruct with an unrolled iterative reconstruction, the input to the regularization stage being an initialization of the unrolled iterative reconstruction.

18. The system of claim 16 wherein the image processor is configured to form the plurality of circular shifted copies by different amounts, the different amounts being different fractions with the factor in a denominator.

19. The system of claim 16 wherein the machine-learned model of the regularization stage comprises a convolutional neural network arranged to receive the input, the input comprising a concatenation of the plurality of circular shifted copies.

20. The system of claim 16 wherein the machine-learned model of the regularization stage comprises a plurality of convolutional neural networks with shared weights independently applied to the plurality of circular shifted copies and a sum and convolution network combining outputs of the convolutional neural networks.

21. A method for reconstruction of a medical image, the method comprising:
scanning, by a magnetic resonance imaging system, a patient using parallel imaging with compressed sensing having an acceleration factor, the scanning resulting in k-space data;
reconstructing, by an image processor, the medical image from the k-space data, the reconstruction comprising a data consistency stage and a regularization stage, an output image of the data consistency stage circularly shifted different amounts to form a plurality of different shifted images, the different amounts based on the acceleration factor, and wherein an input to the regularization stage comprises the plurality of different shifted images or a combination of the plurality of different shifted images; and
displaying the medical image.

Assignees

• Siemens Healthineers Ag

Inventors

• Mostapha Mahmoud
• Körzdörfer Gregor
• Nickel Marcel Dominik
• Raithel Esther
• Arberet Simon
• Nadar Mariappan S

Classifications

• G06T12/20 Primary

  Inverse problem, i.e. transformations from projection space into object space · CPC title

• G06T2211/441

  AI-based methods, deep learning or artificial neural networks · CPC title

• G06T2211/424

  Iterative · CPC title

• G06T2207/10088

  Magnetic resonance imaging [MRI] · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title